Bi-level optimization for the energy conversion efficiency improvement in a photocatalytic-hydrogen-production system

Abstract

Industrialization of the photocatalytic hydrogen production technology is plagued by the low energy conversion efficiency. To address this challenge, this study simultaneously optimizes the structure of the compound parabolic collector (CPC) and the operating conditions of the photoreactor to match the solar flow on the reactor surface and the energy required for chemical reaction at the catalyst interface by building a novel bi-level optimization model with partial differential equation (PDE) constraints. The Sobol method is used to identify the operating parameters that significantly influence the energy conversion efficiency. A new extreme learning machine (ELM) is developed as a surrogate model to approximate the relationship between the operating parameters of the reactor and hydrogen production, reducing computational load. A new nested algorithm is developed to solve the built bi-level optimization model. Numerical results indicate that the new method can find the optimal CPC structure and operating conditions for the photoreactor, achieving an increase in energy conversion efficiency as compared to the results from the systems with un-optimization, optimized CPC structure only and optimized reactor operating parameters only. Our study achieves the synergy of the CPC and photoreactor by combining the advanced optimization technology, machine learning and energy conversion process mechanisms.